Merge pull request !30777 from XianglongZeng/myms_new_2

Merge pull request !30844 from kisnwang/fix-cpu-sparse-ftrl-error

Merge pull request !30813 from yuchaojie/ir_fusion

Merge pull request !30787 from zangqx/zangqx_5

Merge pull request !30783 from 徐安越/master_core

Merge pull request !26841 from 桂宁馨/Sinh_Cosh

Merge pull request !30820 from ZPaC/parse-label-for-parameter-update

Merge pull request !30800 from chenweifeng/priority-replay-buffer

Merge pull request !30769 from Greatpan/dynamic_thread_fix

Merge pull request !30746 from ZPaC/sync-route-table

Merge pull request !30553 from h.farahat/reset

Merge pull request !30799 from yuchaojie/ub_fusion2

Merge pull request !30809 from ling/sr

1. Remove unused fields, such as AnfNode::interpreted_node_, CNode::inputs_value_;
2. Reduce seen_ size from 64 bits to 32 bits;
3. Use bitset for boolean flags;
4. Use pointer for user data;
5. Make some cold fields as user data, such as AnfNode::kernel_info_.

Object size decreased:
AnfNode: 232 -> 144
ValueNode: 264 -> 176
Parameter: 384 -> 272
CNode: 576 -> 400

Merge pull request !30698 from Haim/export_haim

Merge pull request !30654 from huoxinyou/bngrad_nopNode

Merge pull request !30686 from sunsuodong/version

Merge pull request !30589 from hanhuifeng/dcn_dyn_ops_2_new

Merge pull request !30611 from ZPaC/sync-route-table

Merge pull request !30572 from TuDouNi/gpu_cpu_delete_dynamic_kernel

Merge pull request !30614 from huangxinjing/fix_cast_error

Merge pull request !25466 from 桂宁馨/sin_cos